From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from bombadil.infradead.org (bombadil.infradead.org [198.137.202.133]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.lore.kernel.org (Postfix) with ESMTPS id CFCB4C6FD1F for ; Tue, 14 Mar 2023 17:11:46 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; q=dns/txt; c=relaxed/relaxed; d=lists.infradead.org; s=bombadil.20210309; h=Sender: Content-Transfer-Encoding:Content-Type:List-Subscribe:List-Help:List-Post: List-Archive:List-Unsubscribe:List-Id:In-Reply-To:MIME-Version:Message-ID: Subject:Cc:To:From:Date:Reply-To:Content-ID:Content-Description:Resent-Date: Resent-From:Resent-Sender:Resent-To:Resent-Cc:Resent-Message-ID:References: List-Owner; bh=d0P2G+FAU149PJ4OOTwKknKRJigGDfNkIvuwBuZtUog=; b=N4kYTx+xXbgLsO 9+aYeGBLFxK1ne0ZDN7i0S+86Nekfhs+sj+tleJCAL0PQXnhe9R9wsNvzIVuKxQxCSti7HJPn+Q3b upOJXQsptMfsOPreN/s99CreKRmFc7w6wjP8FkCt/aKEoles6w4TnrvxDShxHs05fIL9vlB/Qurjk UTYiIyuUscciLvTVyy500gcw/LF1eVvjaokMQAaIKwB+924sqK4WftHE4ckhW+zp0uhphzLXtuIiD 6XkAUweZlnLe7+UPwdff+hbahwKrwyhuScvEeBfbYNsmKno+vstZHE7KUD8K1baZl/oUOhCLaadoK EkD/gJCT8e/yt0KLeR9w==; Received: from localhost ([::1] helo=bombadil.infradead.org) by bombadil.infradead.org with esmtp (Exim 4.96 #2 (Red Hat Linux)) id 1pc8Ap-00Ax9L-1U; Tue, 14 Mar 2023 17:10:55 +0000 Received: from dfw.source.kernel.org ([139.178.84.217]) by bombadil.infradead.org with esmtps (Exim 4.96 #2 (Red Hat Linux)) id 1pc8Am-00Ax70-0B for linux-arm-kernel@lists.infradead.org; Tue, 14 Mar 2023 17:10:54 +0000 Received: from smtp.kernel.org (relay.kernel.org [52.25.139.140]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by dfw.source.kernel.org (Postfix) with ESMTPS id 708026184A; Tue, 14 Mar 2023 17:10:48 +0000 (UTC) Received: by smtp.kernel.org (Postfix) with ESMTPSA id A4172C433D2; Tue, 14 Mar 2023 17:10:47 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=kernel.org; s=k20201202; t=1678813847; bh=ok1k63vTwaFT1h6WbAsaye8ZP5fcNL1HoRcPKwtbVtI=; h=Date:From:To:Cc:Subject:In-Reply-To:From; b=b5JwnUG5XA7W5kE4jyFISaw2q860l5uvxAHwMuk5lHg+BEEqh0TT2uVLDCxsmCTDQ 3C84R7HuY2lWuRB01w97S61A/z25ENX9FxOXCp7sZTZf+l/2PfEpCR29v8vnqK/fWe IZI1l5Uz3qZCF0f9VqGtsQWMiH7QcvuKHE96fVzJSBRpU0/vOYpFLPc2ANuPmZ36Gm VapqcxYrs+3lnS9Qy5HXZeE9aEb56zmuIvadvnGSg1VyYozj0yLpdCY4Xq9vT0jMie Y6EXwMWAfnQuE1YYVshLMkYLi09aT9Fn5BDqP4nTSxCN7A313gbcyUTl4KJmAJ+Nu3 pY/eTBAqYMyKg== Date: Tue, 14 Mar 2023 12:10:46 -0500 From: Bjorn Helgaas To: Sathyanarayanan Kuppuswamy Cc: Ganapatrao Kulkarni , linux-kernel@vger.kernel.org, linux-pci@vger.kernel.org, linux-arm-kernel@lists.infradead.org, iommu@lists.linux.dev, joro@8bytes.org, bhelgaas@google.com, robin.murphy@arm.com, will@kernel.org, jean-philippe@linaro.org, darren@os.amperecomputing.com, scott@os.amperecomputing.com Subject: Re: [PATCH v2 1/2] PCI/ATS: Add a helper function to configure ATS STU of a PF Message-ID: <20230314171046.GA1655423@bhelgaas> MIME-Version: 1.0 Content-Disposition: inline In-Reply-To: <6d3b4608-f859-f1c3-2391-0017d1b512f4@linux.intel.com> X-CRM114-Version: 20100106-BlameMichelson ( TRE 0.8.0 (BSD) ) MR-646709E3 X-CRM114-CacheID: sfid-20230314_101052_178402_5F60C24D X-CRM114-Status: GOOD ( 55.33 ) X-BeenThere: linux-arm-kernel@lists.infradead.org X-Mailman-Version: 2.1.34 Precedence: list List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Sender: "linux-arm-kernel" Errors-To: linux-arm-kernel-bounces+linux-arm-kernel=archiver.kernel.org@lists.infradead.org On Tue, Mar 14, 2023 at 09:50:06AM -0700, Sathyanarayanan Kuppuswamy wrote: > On 3/14/23 9:02 AM, Bjorn Helgaas wrote: > > On Tue, Mar 14, 2023 at 08:06:07PM +0530, Ganapatrao Kulkarni wrote: > >> On 14-03-2023 06:22 pm, Sathyanarayanan Kuppuswamy wrote: > >>> On 3/14/23 3:08 AM, Ganapatrao Kulkarni wrote: > >>>> On 14-03-2023 04:00 am, Sathyanarayanan Kuppuswamy wrote: > >>>>> On 3/13/23 2:12 PM, Bjorn Helgaas wrote: > >>>>>> On Mon, Feb 27, 2023 at 08:21:36PM -0800, Ganapatrao Kulkarni wrot= e: > >>>>>>> As per PCI specification (PCI Express Base Specification > >>>>>>> Revision 6.0, Section 10.5) both PF and VFs of a PCI EP > >>>>>>> are permitted to be enabled independently for ATS > >>>>>>> capability, however the STU(Smallest Translation Unit) is > >>>>>>> shared between PF and VFs. For VFs, it is hardwired to > >>>>>>> Zero and the associated PF's value applies to VFs. > >>>>>>> > >>>>>>> In the current code, the STU is being configured while > >>>>>>> enabling the PF ATS. Hence, it is not able to enable ATS > >>>>>>> for VFs, if it is not enabled on the associated PF > >>>>>>> already. > >>>>>>> > >>>>>>> Adding a function pci_ats_stu_configure(), which can be > >>>>>>> called to configure the STU during PF enumeration. Latter > >>>>>>> enumerations of VFs can successfully enable ATS > >>>>>>> independently. > > = > >>>>>>> @@ -46,6 +46,35 @@ bool pci_ats_supported(struct pci_dev *dev) > >>>>>>> =A0 } > >>>>>>> =A0 EXPORT_SYMBOL_GPL(pci_ats_supported); > >>>>>>> =A0 +/** > >>>>>>> + * pci_ats_stu_configure - Configure STU of a PF. > >>>>>>> + * @dev: the PCI device > >>>>>>> + * @ps: the IOMMU page shift > >>>>>>> + * > >>>>>>> + * Returns 0 on success, or negative on failure. > >>>>>>> + */ > >>>>>>> +int pci_ats_stu_configure(struct pci_dev *dev, int ps) > >>>>>>> +{ > >>>>>>> +=A0=A0=A0 u16 ctrl; > >>>>>>> + > >>>>>>> +=A0=A0=A0 if (dev->ats_enabled || dev->is_virtfn) > >>>>>>> +=A0=A0=A0=A0=A0=A0=A0 return 0; > >>>>>> > >>>>>> I might return an error for the VF case on the assumption > >>>>>> that it's likely an error in the caller.=A0 I guess one could > >>>>>> argue that it simplifies the caller if it doesn't have to > >>>>>> check for PF vs VF.=A0 But the fact that STU is shared between > >>>>>> PF and VFs is an important part of understanding how ATS > >>>>>> works, so the caller should be aware of the distinction > >>>>>> anyway. > >>>>> > >>>>> I have already asked this question. But let me repeat it. > >>>>> > >>>>> We don't have any checks for the PF case here. That means you > >>>>> can re-configure the STU as many times as you want until ATS > >>>>> is enabled in PF. So, if there are active VFs which uses this > >>>>> STU, can PF re-configure the STU at will? > >>>> > >>>> IMO, Since STU is shared, programming it multiple times is not expec= ted from callers code do it, however we can add below check to allow to pro= gram STU once from a PF. > >>>> > >>>> diff --git a/drivers/pci/ats.c b/drivers/pci/ats.c > >>>> index 1611bfa1d5da..f7bb01068e18 100644 > >>>> --- a/drivers/pci/ats.c > >>>> +++ b/drivers/pci/ats.c > >>>> @@ -60,6 +60,10 @@ int pci_ats_stu_configure(struct pci_dev *dev, in= t ps) > >>>> =A0=A0=A0=A0=A0=A0=A0 if (dev->ats_enabled || dev->is_virtfn) > >>>> =A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 return 0; > >>>> > >>>> +=A0=A0=A0=A0=A0=A0 /* Configured already */ > >>>> +=A0=A0=A0=A0=A0=A0 if (dev->ats_stu) > >>>> +=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 return 0; > >>> > >>> Theoretically, you can re-configure STU as long as no one is using > >>> it. Instead of this check, is there a way to check whether there > >>> are active VMs which enables ATS? > >> > >> Yes I agree, there is no limitation on how many times you write STU > >> bits, but practically it is happening while PF is enumerated. > >> > >> The usage of function pci_ats_stu_configure is almost > >> similar(subset) to pci_enable_ats and only difference is one does > >> ATS enable + STU program and another does only STU program. > > = > > What would you think of removing the STU update feature from > > pci_enable_ats() so it always fails if pci_ats_stu_configure() has not > > been called, even when called on the PF, e.g., > > = > > if (ps !=3D pci_physfn(dev)->ats_stu) > > return -EINVAL; > = > If we are removing the STU update from pci_enable_ats(), why > even allow passing "ps (page shift)" parameter? IMO, we can assume that > for STU reconfigure, users will call pci_ats_stu_configure(). The reason to pass "ps" would be to verify that the STU the caller plans to use matches the actual STU. > Since zero is a valid STU, enabling ATS can be decoupled from STU > update. > > > pci_read_config_word(dev, dev->ats_cap + PCI_ATS_CTRL, &ctrl); > > ctrl |=3D PCI_ATS_CTRL_ENABLE; > > pci_write_config_word(dev, dev->ats_cap + PCI_ATS_CTRL, ctrl); > > = > > Would probably also have to set "dev->ats_stu =3D 0" in > > pci_disable_ats() to allow the possibility of calling > > pci_ats_stu_configure() again. > > = > >> IMO, I dont think, there is any need to find how many active VMs > >> with attached VFs and it is not done for pci_enable_ats as well. > > = > > Enabling or disabling ATS in a PF or VF has no effect on other > > functions. > > = > > But changing STU while a VF has ATS enabled would definitely break any > > user of that VF, so if it's practical to verify that no VFs have ATS > > enabled, I think we should. > = > I also think it is better to check for a ats_enabled status of VF before > configuring the STU. > = > May be something like below (untested), > = > static int is_ats_enabled_in_vf(struct pci_dev *dev) > { > struct pci_sriov *iov =3D dev->sriov; > struct pci_dev *vdev; > = > if (dev->is_virtfn) = > return -EINVAL; > = > for (i =3D 0; i < pci_sriov_get_totalvfs(pdev); i++) { > vdev =3D pci_get_domain_bus_and_slot(pci_domain_nr(dev->b= us), > pci_iov_virtfn_bus(dev, i), > pci_iov_virtfn_devfn(dev, i)= ); I would try hard to avoid pci_get_domain_bus_and_slot(). That's expensive (searches *all* PCI devs with for_each_pci_dev()) and requires dealing with reference counts. Maybe an atomic count in the PF of VFs with ATS enabled. > if (vdev && vdev->ats_enabled) > return 1; > } > = > return 0; > = > } > = > int pci_ats_stu_configure(struct pci_dev *dev, int ps) > { > ... > if (is_ats_enabled_in_vf(dev)) > return -EBUSY; > = > > = > >> Also the caller has the requirement to call either > >> pci_ats_stu_configure or pci_enable_ats while enumerating the PF. > >> > >>>> =A0=A0=A0=A0=A0=A0=A0 if (!pci_ats_supported(dev)) > >>>> =A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 return -EINVAL; > >>>>>> > >>>>>>> + > >>>>>>> +=A0=A0=A0 if (!pci_ats_supported(dev)) > >>>>>>> +=A0=A0=A0=A0=A0=A0=A0 return -EINVAL; > >>>>>>> + > >>>>>>> +=A0=A0=A0 if (ps < PCI_ATS_MIN_STU) > >>>>>>> +=A0=A0=A0=A0=A0=A0=A0 return -EINVAL; > >>>>>>> + > >>>>>>> +=A0=A0=A0 dev->ats_stu =3D ps; > >>>>>>> +=A0=A0=A0 pci_read_config_word(dev, dev->ats_cap + PCI_ATS_CTRL,= &ctrl); > >>>>>>> +=A0=A0=A0 ctrl |=3D PCI_ATS_CTRL_STU(dev->ats_stu - PCI_ATS_MIN_= STU); > >>>>>>> +=A0=A0=A0 pci_write_config_word(dev, dev->ats_cap + PCI_ATS_CTRL= , ctrl); > >>>>>>> + > >>>>>>> +=A0=A0=A0 return 0; > >>>>>>> +} > >>>>>>> +EXPORT_SYMBOL_GPL(pci_ats_stu_configure); > >>>>>>> + > >>>>>>> =A0 /** > >>>>>>> =A0=A0 * pci_enable_ats - enable the ATS capability > >>>>>>> =A0=A0 * @dev: the PCI device > >>>>>>> @@ -68,8 +97,8 @@ int pci_enable_ats(struct pci_dev *dev, int ps) > >>>>>>> =A0=A0=A0=A0=A0=A0=A0=A0=A0 return -EINVAL; > >>>>>>> =A0 =A0=A0=A0=A0=A0 /* > >>>>>>> -=A0=A0=A0=A0 * Note that enabling ATS on a VF fails unless it's = already enabled > >>>>>>> -=A0=A0=A0=A0 * with the same STU on the PF. > >>>>>>> +=A0=A0=A0=A0 * Note that enabling ATS on a VF fails unless it's = already > >>>>>>> +=A0=A0=A0=A0 * configured with the same STU on the PF. > >>>>>>> =A0=A0=A0=A0=A0=A0 */ > >>>>>>> =A0=A0=A0=A0=A0 ctrl =3D PCI_ATS_CTRL_ENABLE; > >>>>>>> =A0=A0=A0=A0=A0 if (dev->is_virtfn) { > >>>>>>> diff --git a/include/linux/pci-ats.h b/include/linux/pci-ats.h > >>>>>>> index df54cd5b15db..7d62a92aaf23 100644 > >>>>>>> --- a/include/linux/pci-ats.h > >>>>>>> +++ b/include/linux/pci-ats.h > >>>>>>> @@ -8,6 +8,7 @@ > >>>>>>> =A0 /* Address Translation Service */ > >>>>>>> =A0 bool pci_ats_supported(struct pci_dev *dev); > >>>>>>> =A0 int pci_enable_ats(struct pci_dev *dev, int ps); > >>>>>>> +int pci_ats_stu_configure(struct pci_dev *dev, int ps); > >>>>>>> =A0 void pci_disable_ats(struct pci_dev *dev); > >>>>>>> =A0 int pci_ats_queue_depth(struct pci_dev *dev); > >>>>>>> =A0 int pci_ats_page_aligned(struct pci_dev *dev); > >>>>>>> @@ -16,6 +17,8 @@ static inline bool pci_ats_supported(struct pci= _dev *d) > >>>>>>> =A0 { return false; } > >>>>>>> =A0 static inline int pci_enable_ats(struct pci_dev *d, int ps) > >>>>>>> =A0 { return -ENODEV; } > >>>>>>> +static inline int pci_ats_stu_configure(struct pci_dev *d, int p= s) > >>>>>>> +{ return -ENODEV; } > >>>>>>> =A0 static inline void pci_disable_ats(struct pci_dev *d) { } > >>>>>>> =A0 static inline int pci_ats_queue_depth(struct pci_dev *d) > >>>>>>> =A0 { return -ENODEV; } > = > -- = > Sathyanarayanan Kuppuswamy > Linux Kernel Developer > = > _______________________________________________ > linux-arm-kernel mailing list > linux-arm-kernel@lists.infradead.org > http://lists.infradead.org/mailman/listinfo/linux-arm-kernel _______________________________________________ linux-arm-kernel mailing list linux-arm-kernel@lists.infradead.org http://lists.infradead.org/mailman/listinfo/linux-arm-kernel